LIGO s Thermal Noise Interferometer: Progress and Status

Transcription

1 LIGO s Thermal Noise Interferometer: Progress and Status Eric Black LSC Meeting Review November 12, 2003 Ivan Grudinin, Akira Villar, Kenneth G. Libbrecht Thanks also to: Kyle Barbary, Adam Bushmaker, Jay Heefner, Fumiko Kawazoe, Luca Matone, Sharon Meidt, Shanti Rao, Kevin Schulz, Michael Zhang Seiji Kawamura

2 Short Term TNI Purpose and Goals Characterize Sapphire for use in Advanced LIGO: Noise Performance, lead time, etc. Contribute to AdvLIGO COC downselect decision. Evaluate thermal noise of coatings for AdvLIGO. Long Term Isolate and study non-gaussian noise in suspensions and mirrors. Reach (and Exceed) the Standard Quantum Limit. Along the way Identify and quantify any unexpected noise sources at sensitivity levels relevant to gravitational-wave detection.

3 TNI Layout

4 TNI Hardware Mode cleaner, arm cavities all contained in one chamber, mounted on a single stack. Wire suspensions similar to LIGO-I Small Optics Suspensions. OSEM actuation, all analog. Laser: 750mW NPRO (Lightwave). All hardware constructed (purchased) and installed.

5 TNI Phase I Expected Spectrum Total Fused-Silica Test Masses Wire Suspensions OSEM Actuation Brownian x ν, m/hz 1/ Radiation Pendulum Thermoelastic Seismic Shot Noise Frequency

6 TNI Phase II Expected Spectrum Total Sapphire Test Masses Wire Suspensions OSEM Actuation x ν, m/hz 1/ Radiation Pendulum Thermoelastic Brownian Seismic Shot Noise Frequency

7 Summer 2001: First data. Schedule: Major Milestones December 2001: SAC noise level approaches scientifically interesting region, ~1e-18 m/ Hz from 1-6 khz, but instrument only locks late at night. NAC noise still high. 2002: Instrument Improvements Improved lock acquisition, stability (robustness). Developing daytime locking capability resulted in loss of sensitivity. Identified, quantified individual noise source contributions. Equalized arm-cavity noise levels. Quality control in photodetectors, beamsplitter; polarization optimization. 2003: Sensitivity Gains, Preliminary Science Results Recovered initial good SAC sensitivity, comparable NAC sensitivity, and retained ability to lock in a challenging seismic environment (MINOS detector assembly next door). Measurement of coating thermal noise on fused silica substrates. March 2004: Provide sapphire data for AdvLIGO COC downselect decision.

8 Latest Sensitivity Curve: 10/23/03 Major improvements since August LSC meeting: Increased modulation depth to suppress photodetector electronic noise. Careful calculation of thermal noise to account for multiple mirrors, difference measurement. Best sensitivity is now approximately 5e-19 m/ Hz. Noise floor appears to be dominated by coating thermal noise from approx. 500 Hz to 20 khz.

9 Calibration δv = DC 1+ DHMC δl To convert the measured voltage into an equivalent length noise, we must know the transfer functions of each servo block. D = Pound-Drever Frequency Discriminant H = Electronic transfer function M = Mirror response (displacement) to signal applied to OSEM controller s FAST-INPUT (voltage) C = Cavity length to frequency conversion factor Specify H and C, verify H by direct measurement. How to determine D and M?

10 Conversion Factor (C) C = ν L = MHz µm This is just math. Converts length change to frequency change. No poles or zeroes, just a number we can calculate. ν = c λ = Hz L =1cm

11 Mirror Response (M): Michelson Measurement Auxilliary Michelson interferometer. Sweep through multiple fringes and count. Rough estimate only!

12 Mirror Response (M): Frequency-Response Measurement Introduce known frequency fluctuation via Laser s PZT. Measure feedback voltage. Well below UGF, transfer function is δf = DH 1+ DHMC δν δν MC Accuracy limited by how well we know the frequency fluctuation δν, which in turn depends on how well we know the laser s PZT response. PZT response measured using reference cavity and known sidebands, ultimately depends on frequency source used to produce sidebands (DS345). Measured PZT response is consistent with manufacturer s spec.

13 Mirror Response (M) Michelson method: Initial rough estimate M 0 () 1 µm V Frequency-response method: Accuracy traceable to commercial frequency reference M 0 ()= µm V Two methods agree. Use frequency-response value for final analysis. NAC and SAC M s are matched to within 7%.

14 Discriminant (D): Sweep Method Direct measurement of Pound-Drever- Hall error signal. Let mirror sweep through a fringe. Identify TEM-00 mode by visibility (not shown in this shot). This is only an estimate of D. D = 2 V t c sb t c 14.75MHz

15 Discriminant (D): UGF Method Measure open-loop transfer function with system locked Fit theory to data New calibration for every lock Blue is data, red is model. One parameter varied for fit: D This method appears to get frequency scaling right. Values obtained are consistent with those obtained by sweep method. Typical: D = 9.0 V MHz Expect: T = DHMC

16 What Causes the Noise Floor? Having established that the instrument is well calibrated, we now want to know the origin of the noise floor. Where in the system does this noise originate? Clues: 1. Incoherence: Differential measurement selects for uncorrelated noise. 2. Scaling with power, modulation depth: True length fluctuations in the arm cavities should be independent of both. Measured voltage should scale linearly with power, product of Bessel functions with modulation depth. Expect δv = RP 0 J 0 ()J β () 1 β 16 F λ δl

17 Power Scaling: δvvs. P 0 Measured voltage noise is proportional to laser power for both high and low modulation depths. Expect this to be true if output voltage is dominated by either Cavity length noise, or Laser frequency noise This measurement rules out the photdetectors or any subsequent measurement electronics as the dominant noise source.

18 Modulation Depth: δvvs. β Discriminant D scales as expected (qualitatively) with modulation depth. Equivalenth length noise in each cavity is independent of modulation depth. Differential noise rises slightly at very low modulation depths due to photodetector electronic noise. Differential noise is only independent of modulation for high-β, where it is quite stable. The noise really is in the cavities, but did it originate there?

19 Servo Noise? Power and modulation-depth scaling indicates that the dominant noise source in the TNI does not come after the cavities in the measurement path. Differential measurement rules out laser frequency noise. Noise must be inside the cavities, but how do we know it originates there? Measure electronic noise going into M, convert to equivalent length noise.

20 Dominant Noise Source is Not Servo Noise

21 What We Have Not Ruled Out 1. OSEM noise: Noise in the local dampers. Mirror transfer function scales as f -2 ; noise floor has f -1/2 dependence. For OSEM noise to be responsible for our noise floor, it would have to have an f +3/2 frequency dependence, which seems unlikely. 2. Thermal noise Pendulum Substrate Coating

22 Pendulum Thermal Noise Pendulum thermal noise theory (above resonant frequency ) S pend ()= f 16k BT 2πf 2 ( ) 5 ω 0 Q pend ω 0 One unknown parameter: Q pend Estimate from similar systems A. Gillespie and F. Raab, Phys. Lett. A, 190, (1994). Music wire with steel clamps, our geometry and loading factor: Q pend

23 Substrate Thermal Noise and Observed Q s Mirror Q s may be determined by localized losses, e.g. Barrel polish (visibly poor) Magnets glued on backs and sides Contact with suspension wires Expect bulk thermal noise to be lower than estimates based on Q measurements Yamamoto, et al., Phys. Lett. A 305, pp (2002) Measured mirror Q s in the TNI range from 1e3 to 3e6. φ substrate = 1 Q < φ substrate < S x () f 2k T B π 32 f 1 ωy φ substrate

24 Q Measurements in Fused Silica Test Masses Q measurements were found to vary from 1700 to over 3 million. Q vs. Frequency for the Fused Silica Mirrors in the TNI 1 E+07 Q (unitless) 1 E+06 1 E+05 1 E+04 NAC Input NAC Output SAC Input SAC Output 1 E Frequency (khz)

25 Coating Thermal Noise Coating thermal noise, best estimate Crooks, et al., Class. Quantum Grav. 19 (5) pp (2002) Harry et al., Class. Quantum Grav. 19 (5) pp (2002) S x ()= f 8k BT π 32 f ~ k BT ωyf 1 ωy φ coating >> ω d φ substrate φ + 1 d Y substrate π ω Y φ + Y Y φ φ substrate + d ω φ coating φ Ringdown coating-q measurements give, but only direct measurement can give? Condition required to distinguish coating thermal noise from substrate thermal noise: For the TNI, ω d 160µm 5µm = 32 φ Expect: φ coating ~ φ substrate Condition met by factor of ten in TNI.

26 TNI Noise Breakdown

27 Next Step: Sapphire Substrates

28 Conclusions TNI noise floor appears to be dominated by coating thermal noise from 500 Hz to 20 khz. Losses in these SiO 2 /Ta 2 O 5 coatings appear to be isotropic: Ringdown measurements give We obtain by experimental fit to our data φ φ φ = φ = TNI is on track for April 2004 AdLIGO COC material downselect. Deadline for thermal noise measurement with fused-silica mirrors was Oct. 29. Our measurement was completed Oct. 23. Current noise floor is below predicted Sapphire noise from 400 Hz to 4 khz. Expect coating thermal noise to go down due to increased Young s modulus of substrate. Will it?